Continuous Estimation of Blood Pressure by Utilizing Seismocardiogram Signal Features in Relation to Electrocardiogram

Abstract

There is an ongoing search for a reliable and continuous method of noninvasive blood pressure (BP) tracking. In this study, we investigate the feasibility of utilizing seismocardiogram (SCG) signals, i.e., chest motion caused by cardiac activity, for this purpose. This research is novel in examining the temporal relationship between the SCG-measured isovolumic moment and the electrocardiogram (PEP_IM). Additionally, we compare these results with the traditionally measured pre-ejection period with the aortic opening marked as an endpoint (PEP_AO). The accuracy of the BP estimation was evaluated beat to beat against invasively measured arterial BP. Data were collected on separate days as eighteen sets from nine subjects undergoing a medical procedure with anesthesia. Results for PEP_IM showed a correlation of 0.67 ± 0.18 (p < 0.001), 0.66 ± 0.17 (p < 0.001), and 0.67 ± 0.17 (p < 0.001) when compared to systolic BP, diastolic BP, and mean arterial pressure (MAP), respectively. Corresponding results for PEP_AO were equal to 0.61 ± 0.22 (p < 0.001), 0.61 ± 0.21 (p < 0.001), and 0.62 ± 0.22 (p < 0.001). Values of PEP_IM were used to estimate MAP using two first-degree models, the linear regression model (achieved RMSE of 11.7 ± 4.0 mmHg) and extended model with HR (RMSE of 10.8 ± 4.2 mmHg), and two corresponding second-degree models (RMSE of 10.8 ± 3.7 mmHg and RMSE of 8.5 ± 3.4 mmHg for second-degree polynomial and second-degree extended, respectively). In the intrasubject testing of the second-degree model extended with HR based on PEP_IM values, the mean error of MAP estimation in three follow-up measurements was in the range of 7.5 to 10.5 mmHg, without recalibration. This study demonstrates the method's potential for further research, particularly given that both proximal and distal pulses are measured in close proximity to the heart and cardiac output. This positioning may enhance the method's capacity to more accurately reflect central blood pressure compared to peripheral measurements.

Keywords: blood pressure; pulse arrival time; pulse transit time; seismocardiogram.

Figure 1

Characteristic points of simultaneously detected signals: invasive BP, ECG, and SCG. IM: isovolumic movement; AO: aortic value opening; AC: aortic valve closure; MO: mitral valve opening.

Figure 2

Distribution of Pearson’s correlation coefficient when PEP_IM and PEP_AO are compared with SBP, DBP, and MAP.

Figure 3

Results of correlation analysis and Bland–Altman plots comparing estimated MAP, based on PEP_IM values using four tested models, and reference MAP measured using IBP.

Figure 4

The average invasively measured blood pressure, heart rate, and pre-ejection period (PEP) changes during the period of the experiment, and the blood pressure estimation made using four models and PEP_IM. All signals are unfiltered and normalized at baseline. The shaded areas indicate the standard deviation. The data represent 18 datasets from 9 subjects (see Table 1). Time = 0 is the time of mannitol infusion. SBP, systolic blood pressure; DBP, diastolic blood pressure. PEP_IM: pre-ejection period measured to IM; PEP_AO: pre-ejection period measured to AO.

Figure 5

Representative plots of the beat-to-beat tracking of the SBP, DBP, and MAP based on invasive BP measurement and estimated BP using PEP_IM-based models #1–#4.

Figure 6

Performance comparison when models’ parameters generated for the first measurement are tested on three follow-up measurements without recalibration. The comparison was made for subjects who underwent four procedure repetitions, namely for Subject 1 and Subject 3 (see Table 1).